Arabic RTL & Japanese vertical books: get direction right first
The hard part of RTL/vertical book translation isn't the language—it's layout direction, punctuation placement, and term alignment. Fixing direction once beats three rounds of rework.
Arabic reads right-to-left; traditional Japanese sets vertically, top-to-bottom, right-to-left. Tools that assume LTR reading order scramble paragraph sequence, punctuation, and digit direction.
A 300-page Arabic medical canon handled through a generic pipeline can take weeks just to re-sequence manually—and still miss Latin names and numbers embedded inside RTL sentences.
Direction detection comes first
Step one is direction detection: automatically identify LTR / RTL / vertical zones and process them separately. Paragraphs, headings, and notes are restored in the correct direction without breaking logical hierarchy.
Proper nouns, numbers, and embedded Latin characters are aligned per target-language rules. Latin drug names and dosage units in Arabic medical texts, for example, must not be flipped by RTL rules.
Native editors check what matters
Once direction and structure are right, native editors focus on punctuation, tone, and cultural fit—not line-by-line re-sequencing. Three weeks, 98.7% direction accuracy, comes from blocking mechanical errors in the automated layer.
Deliverables are direction-correct and structured, ready for InDesign without manual re-layout.